AbstractAn example of simultaneous linear modelling of among and within subject variation through weighted least squares methods is presented as an alternative to standard multivariate analysis of variance techniques. The data pertain to a study of temporal characteristics of some aerodynamic parameters from a speech test used to differentiate groups of individuals with different degrees of palatopharyngeal closure. Both methods are found to yield similar results with respect to the among group comparisons, but the simultaneous strategy enables some further reduction in dimensionality, and thereby extends the description of patterns in the data.